Hongyu Chen

Kelly Marchisio, Saurabh Dash, Hongyu Chen et al.

Quantization techniques are widely used to improve inference speed and deployment of large language models. While a wide body of work examines the impact of quantization on LLMs in English, none have evaluated across languages. We conduct a thorough analysis of quantized multilingual LLMs, focusing on performance across languages and at varying scales. We use automatic benchmarks, LLM-as-a-Judge, and human evaluation, finding that (1) harmful effects of quantization are apparent in human evaluation, which automatic metrics severely underestimate: a 1.7% average drop in Japanese across automatic tasks corresponds to a 16.0% drop reported by human evaluators on realistic prompts; (2) languages are disparately affected by quantization, with non-Latin script languages impacted worst; and (3) challenging tasks like mathematical reasoning degrade fastest. As the ability to serve low-compute models is critical for wide global adoption of NLP technologies, our results urge consideration of multilingual performance as a key evaluation criterion for efficient models.

Wei Zhang, Weiming Zeng, Hongyu Chen et al.

Accurate diagnosis of depression is crucial for timely implementation of optimal treatments, preventing complications and reducing the risk of suicide. Traditional methods rely on self-report questionnaires and clinical assessment, lacking objective biomarkers. Combining fMRI with artificial intelligence can enhance depression diagnosis by integrating neuroimaging indicators. However, the specificity of fMRI acquisition for depression often results in unbalanced and small datasets, challenging the sensitivity and accuracy of classification models. In this study, we propose the Spatio-Temporal Aggregation Network (STANet) for diagnosing depression by integrating CNN and RNN to capture both temporal and spatial features of brain activity. STANet comprises the following steps:(1) Aggregate spatio-temporal information via ICA. (2) Utilize multi-scale deep convolution to capture detailed features. (3) Balance data using the SMOTE to generate new samples for minority classes. (4) Employ the AFGRU classifier, which combines Fourier transformation with GRU, to capture long-term dependencies, with an adaptive weight assignment mechanism to enhance model generalization. The experimental results demonstrate that STANet achieves superior depression diagnostic performance with 82.38% accuracy and a 90.72% AUC. The STFA module enhances classification by capturing deeper features at multiple scales. The AFGRU classifier, with adaptive weights and stacked GRU, attains higher accuracy and AUC. SMOTE outperforms other oversampling methods. Additionally, spatio-temporal aggregated features achieve better performance compared to using only temporal or spatial features. STANet outperforms traditional or deep learning classifiers, and functional connectivity-based classifiers, as demonstrated by ten-fold cross-validation.

Gongning Luo, Mingwang Xu, Hongyu Chen et al.

Breast cancer is one of the most common causes of death among women worldwide. Early detection helps in reducing the number of deaths. Automated 3D Breast Ultrasound (ABUS) is a newer approach for breast screening, which has many advantages over handheld mammography such as safety, speed, and higher detection rate of breast cancer. Tumor detection, segmentation, and classification are key components in the analysis of medical images, especially challenging in the context of 3D ABUS due to the significant variability in tumor size and shape, unclear tumor boundaries, and a low signal-to-noise ratio. The lack of publicly accessible, well-labeled ABUS datasets further hinders the advancement of systems for breast tumor analysis. Addressing this gap, we have organized the inaugural Tumor Detection, Segmentation, and Classification Challenge on Automated 3D Breast Ultrasound 2023 (TDSC-ABUS2023). This initiative aims to spearhead research in this field and create a definitive benchmark for tasks associated with 3D ABUS image analysis. In this paper, we summarize the top-performing algorithms from the challenge and provide critical analysis for ABUS image examination. We offer the TDSC-ABUS challenge as an open-access platform at https://tdsc-abus2023.grand-challenge.org/ to benchmark and inspire future developments in algorithmic research.

Hongyu Chen, Yiqi Gao, Min Zhou et al.

Recently, integrating visual controls into text-to-image~(T2I) models, such as ControlNet method, has received significant attention for finer control capabilities. While various training-free methods make efforts to enhance prompt following in T2I models, the issue with visual control is still rarely studied, especially in the scenario that visual controls are misaligned with text prompts. In this paper, we address the challenge of ``Prompt Following With Visual Control" and propose a training-free approach named Mask-guided Prompt Following (MGPF). Object masks are introduced to distinct aligned and misaligned parts of visual controls and prompts. Meanwhile, a network, dubbed as Masked ControlNet, is designed to utilize these object masks for object generation in the misaligned visual control region. Further, to improve attribute matching, a simple yet efficient loss is designed to align the attention maps of attributes with object regions constrained by ControlNet and object masks. The efficacy and superiority of MGPF are validated through comprehensive quantitative and qualitative experiments.

Hongcheng Gao, Zihao Huang, Lin Xu et al.

Achieving human-like perception and reasoning in Multimodal Large Language Models (MLLMs) remains a central challenge in artificial intelligence. While recent research has primarily focused on enhancing reasoning capabilities in MLLMs, a fundamental question persists: Can Multimodal Large Language Models truly perceive the world as humans do? This paper shifts focus from reasoning to perception. Rather than constructing benchmarks specifically for reasoning, we introduce the Turing Eye Test (TET), a challenging perception-oriented benchmark comprising four diagnostic tasks that evaluate MLLMs' performance on synthetic images that humans process intuitively. Our findings reveal that state-of-the-art MLLMs exhibit catastrophic failures on our perceptual tasks trivial for humans. Both in-context learning and training on language backbone-effective for previous benchmarks-fail to improve performance on our tasks, while fine-tuning the vision tower enables rapid adaptation, suggesting that our benchmark poses challenges for vision tower generalization rather than for the knowledge and reasoning capabilities of the language backbone-a key gap between current MLLMs and human perception. We release a representative subset of TET tasks in this version, and will introduce more diverse tasks and methods to enhance visual generalization in future work.

Jerry Yao-Chieh Hu, Hude Liu, Hong-Yu Chen et al.

We prove that with linear transformations, both (i) two-layer self-attention and (ii) one-layer self-attention followed by a softmax function are universal approximators for continuous sequence-to-sequence functions on compact domains. Our main technique is a new interpolation-based method for analyzing attention's internal mechanism. This leads to our key insight: self-attention is able to approximate a generalized version of ReLU to arbitrary precision, and hence subsumes many known universal approximators. Building on these, we show that two-layer multi-head attention alone suffices as a sequence-to-sequence universal approximator. In contrast, prior works rely on feed-forward networks to establish universal approximation in Transformers. Furthermore, we extend our techniques to show that, (softmax-)attention-only layers are capable of approximating various statistical models in-context. We believe these techniques hold independent interest.

Hongyu Chen, Bingliang Jiao, Wenxuan Wang et al.

Lifelong person re-identification attempts to recognize people across cameras and integrate new knowledge from continuous data streams. Key challenges involve addressing catastrophic forgetting caused by parameter updating and domain shift, and maintaining performance in seen and unseen domains. Many previous works rely on data memories to retain prior samples. However, the amount of retained data increases linearly with the number of training domains, leading to continually increasing memory consumption. Additionally, these methods may suffer significant performance degradation when data preservation is prohibited due to privacy concerns. To address these limitations, we propose using textual descriptions as guidance to encourage the ReID model to learn cross-domain invariant features without retaining samples. The key insight is that natural language can describe pedestrian instances with an invariant style, suggesting a shared textual space for any pedestrian images. By leveraging this shared textual space as an anchor, we can prompt the ReID model to embed images from various domains into a unified semantic space, thereby alleviating catastrophic forgetting caused by domain shifts. To achieve this, we introduce a task-driven dynamic textual prompt framework in this paper. This model features a dynamic prompt fusion module, which adaptively constructs and fuses two different textual prompts as anchors. This effectively guides the ReID model to embed images into a unified semantic space. Additionally, we design a text-visual feature alignment module to learn a more precise mapping between fine-grained visual and textual features. We also developed a learnable knowledge distillation module that allows our model to dynamically balance retaining existing knowledge with acquiring new knowledge. Extensive experiments demonstrate that our method outperforms SOTAs under various settings.

Arash Ahmadian, Saurabh Dash, Hongyu Chen et al.

Emergent properties have been widely adopted as a term to describe behavior not present in smaller models but observed in larger models. Recent work suggests that the trade-off incurred by quantization is also an emergent property, with sharp drops in performance in models over 6B parameters. In this work, we ask "are quantization cliffs in performance solely a factor of scale?" Against a backdrop of increased research focus on why certain emergent properties surface at scale, this work provides a useful counter-example. We posit that it is possible to optimize for a quantization friendly training recipe that suppresses large activation magnitude outliers. Here, we find that outlier dimensions are not an inherent product of scale, but rather sensitive to the optimization conditions present during pre-training. This both opens up directions for more efficient quantization, and poses the question of whether other emergent properties are inherent or can be altered and conditioned by optimization and architecture design choices. We successfully quantize models ranging in size from 410M to 52B with minimal degradation in performance.

Imanol Luengo, Maria Grammatikopoulou, Rahim Mohammadi et al.

Surgical scene segmentation is essential for anatomy and instrument localization which can be further used to assess tissue-instrument interactions during a surgical procedure. In 2017, the Challenge on Automatic Tool Annotation for cataRACT Surgery (CATARACTS) released 50 cataract surgery videos accompanied by instrument usage annotations. These annotations included frame-level instrument presence information. In 2020, we released pixel-wise semantic annotations for anatomy and instruments for 4670 images sampled from 25 videos of the CATARACTS training set. The 2020 CATARACTS Semantic Segmentation Challenge, which was a sub-challenge of the 2020 MICCAI Endoscopic Vision (EndoVis) Challenge, presented three sub-tasks to assess participating solutions on anatomical structure and instrument segmentation. Their performance was assessed on a hidden test set of 531 images from 10 videos of the CATARACTS test set.

Hongyu Chen, Zhijie Yang, Xiting Zhao et al.

This paper presents a fully hardware synchronized mapping robot with support for a hardware synchronized external tracking system, for super-precise timing and localization. Nine high-resolution cameras and two 32-beam 3D Lidars were used along with a professional, static 3D scanner for ground truth map collection. With all the sensors calibrated on the mapping robot, three datasets are collected to evaluate the performance of mapping algorithms within a room and between rooms. Based on these datasets we generate maps and trajectory data, which is then fed into evaluation algorithms. We provide the datasets for download and the mapping and evaluation procedures are made in a very easily reproducible manner for maximum comparability. We have also conducted a survey on available robotics-related datasets and compiled a big table with those datasets and a number of properties of them.

Minghan Yang, Dong Xu, Hongyu Chen et al.

In this paper, we consider stochastic second-order methods for minimizing a finite summation of nonconvex functions. One important key is to find an ingenious but cheap scheme to incorporate local curvature information. Since the true Hessian matrix is often a combination of a cheap part and an expensive part, we propose a structured stochastic quasi-Newton method by using partial Hessian information as much as possible. By further exploiting either the low-rank structure or the kronecker-product properties of the quasi-Newton approximations, the computation of the quasi-Newton direction is affordable. Global convergence to stationary point and local superlinear convergence rate are established under some mild assumptions. Numerical results on logistic regression, deep autoencoder networks and deep convolutional neural networks show that our proposed method is quite competitive to the state-of-the-art methods.

Jianxiong Cai, Jiawei Hou, Yiren Lu et al.

In this paper, we focus on the question: how might mobile robots take advantage of affordable RGB-D sensors for object detection? Although current CNN-based object detectors have achieved impressive results, there are three main drawbacks for practical usage on mobile robots: 1) It is hard and time-consuming to collect and annotate large-scale training sets. 2) It usually needs a long training time. 3) CNN-based object detection shows significant weakness in predicting location. We propose an improved method for the detection of planar objects, which rectifies images with geometric information to compensate for the perspective distortion before feeding it to the CNN detector module, typically a CNN-based detector like YOLO or MASK RCNN. By dealing with the perspective distortion in advance, we eliminate the need for the CNN detector to learn that. Experiments show that this approach significantly boosts the detection performance. Besides, it effectively reduces the number of training images required. In addition to the novel detection framework proposed, we also release an RGBD dataset and source code for hazmat sign detection. To the best of our knowledge, this is the first work of image rectification for CNN-based object detection, and the dataset is the first public available hazmat sign detection dataset with RGB-D sensors.

Hongyu Chen, Sören Schwertfeger

Many robotics and mapping systems contain multiple sensors to perceive the environment. Extrinsic parameter calibration, the identification of the position and rotation transform between the frames of the different sensors, is critical to fuse data from different sensors. When obtaining multiple camera to camera, lidar to camera and lidar to lidar calibration results, inconsistencies are likely. We propose a graph-based method to refine the relative poses of the different sensors. We demonstrate our approach using our mapping robot platform, which features twelve sensors that are to be calibrated. The experimental results confirm that the proposed algorithm yields great performance.

Hongyu Chen, Xiting Zhao, Jianwen Luo et al.

This paper presents a fully hardware synchronized mapping robot with support for a hardware synchronized external tracking system, for super-precise timing and localization. We also employ a professional, static 3D scanner for ground truth map collection. Three datasets are generated to evaluate the performance of mapping algorithms within a room and between rooms. Based on these datasets we generate maps and trajectory data, which is then fed into evaluation algorithms. The mapping and evaluation procedures are made in a very easily reproducible manner for maximum comparability. In the end we can draw a couple of conclusions about the tested SLAM algorithms.